607 
0361-0365/98/1200-0607$15.00/0  q   1998 Hum an Sciences Press, Inc. 
R esearch in H igher E ducation, V ol. 39, No. 6, 1998 
DETERMINANTS OF RESEARCH
PRODUCTIVITY IN HIGHER EDUCATION 
Halil Dundar and Darrell R. Lewis 
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 
As research expenditures have risen and as sources of research funding have
changed, an increased emphasis on research performance has developed in U.S.
research universities. Although much of the historical debate has centered around
the individual attributes of faculty, several recent studies have begun to focus on the
effect of program or organizational factors as  powerful attributes for enhancing such
productivity. This paper extends the findings of these recent studies by examining
the relationship between academ ic research productivity and institutional factors
from the most recent National Research Council data on the nation’s research uni-
versities and their programs in the four broad fields of the biological sciences, engi-
neering, the physical sciences and mathematics, and the social and behavioral sci-
ences. Several findings are recommended for institutional policymakers. 
: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : 
W hat ar e the determina nts of  the r esear ch per formance of doctora l programs
in A merica n r esearc h unive rsities? A lthough this que stion appea rs to be rela-
tively easy to answ er, surpr isingly it has not be en fully examine d. F ortuna te ly,
the most r ecent survey of the  quality assessment of doctoral pr ogr am s by the
N ationa l Resear ch C ouncil (G oldberger e t al., 1995)  has provided us w ith a set
of da ta  that permits such e xa m ination. T his pa per estim a te s se ve ral r egression
models to exa m ine the possible pr edic tors and organizational fac tor s influenc-
ing r ese arc h per formance in U .S. doc tor al- level institutions.
Par ticipants in A m er ic an higher  educa tion have long been intere sted in un-
covering those  polic y and institutional f actors that not only assoc ia te  w ith but
also fac ilita te  enhanced pr oduc tivity in both resea rch a nd instr uctional  a ctivities
in re sea rch-level institutions. A lthough a  large  num ber  of studie s have bee n
unde rtaken to r ate and rank graduate-level programs, beginning w ith H ughes
(1925) in his early study of graduate schools and continuing up thr ough similar 
H alil D undar, Resear ch A ssociate of Planning, U niversity of M innesota; D arr ell R. Lew is, P ro-
fessor of E ducational P olicy, U niversity of  Minnesota. A ddress corr espondence to: D ar rell R.
L ew is, D epar tm ent of E ducational Policy and A dm inistration, 310H  Wulling H all, U niversity of
Minnesota, Minneapolis, MN  55455; e-m ail: lew is004 6 m aroon.tc.umn.edu.

608DUNDAR  AND LEWIS 
w ork by Car tter  (1966) , Roose and A nderson ( 1970), Jones, L indzey, and
Coggeshall ( 1982), and the most rec ent r evie w s by G oldberger, Ma he r, and
F la ttau (1995), none of  the se  studie s a ttempte d to direc tly address and examine
the predictors a nd f actors influencing a ctual per formances.
A lthough re putational studies ha ve  r eceived a great deal of attention w ithin
highe r educa tion, debate continues as to how  r elia ble they a re a s a  m e asur e of a
progra m 's a ctua l productivity. A ccor dingly, a number of studies have searc he d
for  fa ctor s tha t corre la te d w ith progra m  ratings ( e.g., H agstrom , 1971; D rew
and K arpf , 1981; L aband, 1985; Saunier, 1985; Webster, 1986; Fairw e ather,
1988; C onr ad and B la ckbur n, 1985; Young et al., 1987; Tan, 1990, 1992; Taut-
koushian et a l., in pre ss). E ve n though the re are  som e  conc erns a bout the
methods and sam ple popula tions e m ploye d in these studies, most of the studies
tend to confirm tha t pr ogr am  ratings ar e indeed highly corr elated w ith only a
fe w  f actors: T he tw o most important a ppea red to be program size and the  r e-
sear ch productivity of individua l fac ulty. E ven though most of these studies
have ide ntified some c or relates w ith program ratings, it is still largely an un-
answ ere d que stion as to w ha t an institution c an do to help inc rease its program
quality a nd its resea rch productivity. A s a re sult, the purpose of this paper is to
examine the  resear ch pr oduc tivity of doctoral-gra nting universities as m e asur ed
through the  most r ecent survey sponsore d by the  N ationa l Resea rch C ounc il,
and to explore  those departmenta l a nd institutiona l f actors that contribute to
such m ea sure s of productivity.
A lthough productivity in higher  education has an obvious multidimensional
char acter as it r elates to both know ledge pr oduction and know ledge dissemina tion
through its various f orm s of resear ch, tea ching, and outre ach ac tivities, resear ch
productivity in pa rtic ula r has re ceived a gr eat a m ount of  attention and c once rn.
Re se arch effort and output f or m  a ver y distinguishing part of  the def initional
char acter of Amer ic an re sea rch universities and, as a conse quenc e, the public
ra nkings of  a cademic programs have be com e incre asingly impor ta nt. Spurre d by
college ra nkings in the  popular pr ess, numerous studie s have rated the  nation's
colleges and universities base d on sever al subjec tive (e.g., judgments about the
quality of program faculty or the qua lity of  their teac hing) and objective m e asur es
(e .g., r esearc h per formance, student selec tivity, a nd the  like). In 1995, f or  e xa m -
ple, w he n the  N ational R esearc h Council released its la test comprehensive study
examining the ra tings of  doc tora l pr ogr am s of the na tion's r esearc h universities
(G oldberger et al., 1995), se ve ral studies investigated the corre la te s of these
ra tings and f ound, as expe cted, that the  resear ch produc tivity of the a cademic
progra m s w as highly related to their favorable reputation ( e.g., E hr enberg and
H urst, 1996; Tautkoushian e t al., in pr ess). N one of these rece nt studies a ttempte d
to dir ectly infe r ca usation, only a ssociation. It w as noted that re putation f or
schola rly excellence ca n, in turn, result in an increa se d c apacity for attrac ting
re se arch and high- ability gra duate stude nts to the program (G runig, 1997).

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION609 
O n the other  hand, som e institutional adm inistrators have com plained that
their  sm a ll pr ogr am  siz e undoubtedly contributed to their low  reputation, but
that such measures did not ac curately acc ount for  either  the ir  distinguishing
high quality or  the str ength of the ir  a ve rage fa culty produc tivity. T hey a rgued
that such r eputa tional r ankings w er e he avily inf luence d sim ply by the siz e of
large  pr ogr am s inde pe nde nt of the qua lity of  average  f aculty productivity. T his
view  of the linkages betw een pr oductivity, orga niza tional siz e, and reputation is
not w ell under stood. I n othe r w ords, the policy question bec omes w hether an
increa se in the num ber  of f aculty membe rs w ill, by itself, re sult in an improve-
ment in pr ogr am pr oductivity. If so, how  a nd to w ha t e xtent?
T he pr esent study adds to our know le dge about r esearc h productivity in
higher  education in se ve ral im portant w ays that inc lude both new  data a nd new
models w ith better mode l specifica tion and better  m e asur eme nt of f actors. F ir st,
the study uses the most rec ent 1995 N RC da ta  and updates previous studies
w ith an expanded and respecified  m odel. I t e xtends this pre vious w or k by intro-
ducing new  measures of influencing f actors. Se cond, it examines the data by
field cluste rs (i.e ., soc ial science s, engineering, biologica l sc ienc es, physica l
sciences), ra the r than by departmenta l for m s ( i.e., e conomics, ele ctrical engi-
neer ing, chem istry) or institutional-level data. S uch an aggre ga te d da ta  se t of
field-re late d clusters provide s us w ith a  stronger sample and grea te r confidence
of predictability w ith our  regre ssion mode ls. 
THE LITERATURE ON RESEARCH PRODUCTIVITY 
Productivity studie s in highe r education have been incre asing in importance
since  the ear ly 1970s. N um er ous and diff ering types of studies have examined
fa ctor s a ff ecting the productivity of  universities and their individua l a cademic
programs or  the ir  fac ulty (see, for exa m ple, G r aves et al., 1982; Clark and
L ew is, 1985; C resw ell, 1985; G olde n et al., 1986; John, 1994; Johnes, 1988b;
L ong, 1978; L evin and Stephan, 1989; M eador et al., 1992; O lson, 1994) . D e -
spite the  prolifer ation of  such studie s, ver y f ew  have  addre ssed the  par ticular
problem  tha t is related to the  multiproduc t nature of m ost univer sitie sÐthat is,
the production a nd com bination of  teaching, r esearc h, a nd ser vic e/outrea ch ac -
tivitie s. O nly a  f ew  studie s, for example, ha ve addr esse d ec onomie s of scale
and scope in the context of  the  multidimensional outc om es of univer sitie s
(C ohn et al., 1989; L ew is and D undar, 1995). A s a consequence, m ost produc-
tivity studies in highe r education have tended to focus on instruc tion a nd
large ly on unde rgra duate instr uc tion. N ever the less, sever al studies have at-
tem pted to e xa m ine those institutional fac tor s that c ontribute  to resea rch pro-
ductivity. Ma ny have perce ived tha t measuring re sea rch perfor m anc e is a  r ela-
tively ea sy task bec ause  of  rea dily availa ble measures suc h a s published books,
jour nal ar ticles, or c itation counts a cross univer sitie s.  D espite  the a va ilability  of

610DUNDAR  AND LEWIS 
such outcom e  measure s, studies examining re se arch productivity still have  r e-
maine d quite  lim ite d. T his largely has bee n due to the  measurement problem s
of both r esearc h inputs and outputs.
T he unit of analysis f or resear ch productivity ca n be at individual, de part-
menta l or  c ollegiate, and institutional levels. Yet most of the ear ly productivity
studies of fa culty re sea rch per formance f oc used on individua l f aculty r athe r
than at the depar tmental or  institutional le ve ls. 
Individual Attributes 
Individual fac ulty produc tivity studies have examine d a w ide  range of f actors
affe cting fac ulty resear ch pr oduc tivity. E arly w ork w ith the se f actors included
the eff ects of age, gender, soc ioe conomic status, and educational background
(B ell a nd S eater, 1980; B raxton and Ba ye r, 1986; Clark and L ew is, 1985;
Cr esw ell, 1986; L evin and Stephan, 1989; L ew is and B ecker, 1979; Tie n and
Blac kbur n, 1996), a long w ith seve ral cultural and organizational dimensions
(C onr ad and B la ckbur n, 1986) . A  pa rtic ular tracking relationship, for  example,
has bee n betw een age, experience, and productivity. A lthough the re appears to
be a str ong age-  a nd experienc e-productivity r elationship in econom ic theory
(i.e., as age and exper ienc e inc rease, productivity a lso inc reases up to a point
and then appe ars to level off ), this re la tionship ha s bee n f ound to be more
mixed in highe r e duca tion and var ies by f ield ( Clark and L e w is, 1985; L evin
and S te phe n, 1989) . Ne vertheless, it ha s also bee n noted that gener ally f ull a nd
more se nior professors (par ticularly at r ese arc h univer sities) tend to ha ve a ccu-
mulative a dvantages over m ost assista nt a nd associate prof essors that re sult in
highe r levels of  productivity ( Cole and Cole, 1972; Clar k and L ew is, 1985;
L ong, 1978) .
T hese fac tor s gener ally have bee n introduc ed in severa l differe nt w ays. M ost
have included individua l char acteristic s, f or  exam ple, innate attr ibutes such as
ability, stam ina, per sonality, ge nde r, a ge , and yea rs of experience ( Cre sw ell,
1986) . O ther individual a ttributes have  be en related to e nvironmenta l location,
including quality of  gra dua te  training, pre stige  of employing department or in-
stitution, com munic ation netw orks, and fr eedom  in the w orkplace (see , f or e x-
ample , Cole and Cole, 1972).
T he culture of  a depa rtm ent or institution also has been f ound to be a n im-
portant fa ctor  determining r ese arc h perf or m anc e of individual f aculty. C ulture
re la te s to share d attitudes and values in a n aca de m ic unit. A  r esear ch-oriented
cultur e exists w hen all faculty and adm inistrators are  socialized to be str ong
re se archers during their graduate tra ining, value resear ch, and m a intain contin-
uous inte rnal a nd external comm unic ation w ith other  r esearc hers, and hire  ne w
fa culty w ith str ong re se arch cr edentia ls (Cr esw ell, 1986).

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION611 
Departmental and Institutional Attributes 
More  useful for public and institutional polic y purposes have been studies
that foc us la rgely on de pa rtm e ntal and institutiona l attributes that m ight relate
to e nha nc ed resear ch pr oductivity. O ne  c ritica l dim ension of depar tmental r e-
searc h productivity is its relationship to organizational and fa culty size. A
grow ing num ber  of  studies (e.g., Be ll and S eater, 1980; Rushton and M eltzer,
1981; Baird, 1986, 1991; C rew e, 1988; Jordan et al., 1988, 1989; G olden et a l.,
1986, 1992a, 1992b, Johnson et al., 1995) ha ve  dir ectly e xa m ined fac ulty size ,
most rece ntly the study conducted by K yvik ( 1995).
K yvik ide ntified several argum ents in fa vor  of  larger  departm e nta l size for
re sea rch productivity. F irst, la rger  de partm e nts can better f acilitate collaborative
re sea rch groups. In larger de pa rtm ents the re ar e more likely to be sever al fac-
ulty w ith simila r r ese arch inter ests, w hich may incre ase cooper ation  and c ollab-
oration f or  joint r ese arc h products. Q uite simply, resear ch perfor m ance  is likely
to incr ease as a re sult of  gr eater  intera ction betw een department members. T his
process ha s been re fer red to as ªinte llec tual synergyº (K yvik, 1995). S econd,
large r de pa rtm e nts a re m ore  like ly to attract higher-quality resear chers. Finally,
large r de pa rtm e nts may ha ve  gre ater amounts of resources w ith more degree s of
fr eedom  in their use. D espite these pote ntial a dva ntages of la rge department
siz es for resear ch perf orm anc e, it has a lso been argued that there  are  several
disadvantages to large  siz e. A s a depar tment's size  inc rea ses, resea rch per for-
mance can be hampe red due to increa sing difficulty in c omm unication and
mor e f orm a l rules and r outines tha t m a y hinder  initiative s and innovative ness.
Mor eover, a s em e rging te chnologies provide m ore  and more opportunities for
fa culty to c ollaborate w ith their collea gue s at othe r universities w orldw ide, a
large  depa rtm ent m ay not provide the assumed benefit of  c ollaborations for
suc h r esearc h.
In fac t, m uch of  the r esearc h on f aculty size is mixed. Some have a rgued tha t
w hat m a ny ha ve  m e asur ed or estimated is sim ply total depar tmental resea rch
production and not quality or aver age fac ulty pr oductivity. I n an ea rly study of
the deter m inants of  resea rch pr oductivity, Blackburn a nd his c olleagues (1978)
re por te d that de pa rtm e nt size  wa s a r elatively poor pre dic tor of r ese arch pro-
ductivity as m ea sure d by the total num ber of publications per fac ulty member.
It w as e stim ated that beyond a depar tmental size  of 11 to 15 faculty m e m ber s,
productivity r em a ined r ela tively stable in their study.
K yvik (1995) examine d the r elationship betw een size of depar tm ents and
re sea rch produc tivity in N orw ay's four universities in five fields of study (i.e .,
in the humanities, social science s, natura l sciences, m e dica l sciences, and te ch-
nology). R ese arc h productivity w as measured as the number of publications
betw een 1989 and 1991. T he public ations w ere journal ar ticles, a rtic le s in re-

612DUNDAR  AND LEWIS 
sear ch books, textbooks and c onf ere nce pr oc eedings, resea rch books publishe d
by book companie s, and reports published in r eport se ries. A  single  index of
re se arch productivity wa s cr eated using a w eighting syste m . U sing P earson's
cor relation analysis, K yvik reported no signif ica nt relationships betw ee n size
and resear ch productivity exce pt in the natur al sciences. Ma rtin and Skea
(1992) also found no re la tionship betw een siz e and produc tivity in a study of
Br itish univer sity depar tments expressly in the natura l sc ie nces f ie ld. H ow ever,
they did f ind a significa nt re la tionship  w ith subf ie ld groups w ithin a  de pa rtm e nt
w hen the faculty w ere  able to share  some of  the same r ese arc h equipment or
inte ract closely w ith e ach other.
Jor da n and his colle agues (1988, 1989), on the other hand, e xa m ined the
re la tionship  be tw ee n  ac ademic  depar tmental size and public ations  in the U nited
S ta tes a nd f ound a positive re la tionship. D e pa rtm ental productivity w as as-
sum ed to be measured by avera ge  fac ulty publication w ithin a given depart-
ment. T hey f ound that departmenta l r esear ch produc tivity w a s c losely related to
progra m  size as m e asur ed by the numbe r of  program f aculty, but that depart-
menta l resear ch produc tivity incre ases only up to a  point in size a nd then sta rts
to decline as the numbe r of fac ulty increa ses beyond som e point of scale.
G olde n a nd Ca rste nsen (1992a) re exam ined the  same data and reported tha t the
effe ct of depar tmental size  w as even sm a ller  and actually de clines a fter c ontrol-
ling for sever al other  fa ctor s. A fter  several itera tions betw e en these  tw o groups
in the re se arch liter ature,  the re sults w ere  still la rgely inconclusive due to prob-
lems in both the  measurement of variables a nd in specifying the  estim ation
mode l.
Cre w e ( 1988) examined resear ch productivity in the depa rtm ents of politics
in U .K . universities a nd found a la rge variation in departmenta l aver age pub-
lication r ates. H e suggested tha t these diff erenc es r esulted fr om  diff erenc es in
re sourc es and opportunities for re se arch. Such possible  factors as leave entitle-
ments, travel money, te aching loa ds, the  availability of r esearc h funds, and the
re se arch e thos in a  department w e re a ll exa m ined and off ered a s plausible ex-
planations.
In shor t, depar tmental size of te n is seen a s a  critic al fac tor in f acilitating
re se arch. L arge  de pa rtm ents may simply become mor e powe rful w ithin a  col-
lege or  university a nd r ece ive m ore facilita ting resources f or  resea rch ac tiv-
itie s Ðsuch as equipm ent, supplies, sec retar ial suppor t, re sea rch assistants,
trave l f unds, or tea ching re pla cements for  those on le aveÐand these resource s
may fac ilita te  greater  resear ch perf or m anc e. T he  accumulative effec ts of f aculty
siz e ar e im portant. In a ddition, de pa rtm e ntal ec onomie s of sca le  m ay ar ise a nd
lead to even m ore  e ff ic ienc y through share d use of such r esour ces.
Institutional control ( i.e., pr ivate versus public) also has bee n an are a r eport-
ing m ixed r esults in previous r esearc h. Jordan a nd his colleagues (1988, 1989)
examined the  effects of type of organizational influence a nd contr ol ( i.e., public

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION613 
or private) on departmental r esearc h pr oductivity. T hey reportedly f ound str ong
evidence that private institutions w ere associated w ith grea ter  ac adem ic re-
searc h productivity. H ow eve r, in a reana lysis of  the  sam e  data set, G olden and
Ca rstensen ( 1992b) reported that the e ffe ct of institutional  control decline s after
controlling for both r esearc h support and the de partm e nt's r eputational rating.
T hey a rgued that this finding is consistent w ith the view  tha t departments in
private institutions em phasize r esearc h over te aching and service activitie s,
w hile de partm e nts in public univer sitie s give gre ater e m phases to tea ching,
public se rvice, and outrea ch. In their w ords, ªPr ivate institutions may not be
mor e efficient in their r esour ce use than are  public unive rsities; the latter may
produce more tea ching and service outputs per faculty m ember, pr ovide fe w er
support fac ilitie s and pay low er  salar ie sº ( p. 160) .
Beyond the size of  fa culties and organizational contr ol, othe r depa rtm ental
variables have be en found to be at lea st cor related w ith de partm e ntal r esearc h
perf orm ance. T hese factors ha ve  include d the annua l resea rch spending of the
department, the num ber  of students in the de pa rtm ent, and the perc enta ge of
departmental f aculty holding resea rch grants (see, f or example , G runig, 1997).
In the context of  British exper ienc es with departmenta l r esearc h pe rformance ,
Johnes (1988a)  noted that the student±sta ff  r atio, the quality of com puting facil-
ities, the size  of  the libr ary, and the availability of  secr etarial, a dministr ative ,
and teaching a ssistance w e re all factors tha t might inf luenc e the r esearc h per-
formance  of a depar tment. H e  also noted that a nother possible explana tion of
diffe rences in r esearc h output m ight be the qua ntity of nongove rnme nt resea rch
funding acquired by the univer sity. A  number of othe r studie s also have focused
on the use of te chnology to enhance  both instructional  and r esear ch productivity
(e .g., Massy and Wilger, 1995).
Institutions play a significa nt r ole  in determining both individual and depart-
menta l produc tivity. De spite the importance of institutional fa ctor s on re se arch
perf orm ance, very f ew  such studies exist due  in part to the lack of appropriate
output da ta  at the institutiona l level and to measurement problems ac ross insti-
tutions. Yet se ve ral attem pts have been made to examine  institutional re se arch
perf orm ance. S evera l studies indicated tha t publication patterns of the  f aculty
may va ry from institution to institution because of diff erences in institutional
norms. Long (1978), f or  exam ple, found that w he n natural scientists m oved to a
new  institution, the ir  pr oduc tion patter ns soon r eflec te d the publica tion nor m s
of the  new  institution. B entley and B la ckbur n ( 1990) analyzed changes in insti-
tutional re se arch perf or m anc e ( publication  and gr ant support) betw een 1979 a nd
1988 and found tha t there w er e substantial cha nges in institutional perf or m anc e
over tim e. T hey also noted that ªacc umula tive advantageº a nd ªr esour ces and
re cognitionº contribute to resear ch suppor t and in turn le ad to gre ater resear ch
productivity.
In another institutional-level study, Rushton and M eltzer (1981)  examine d

614DUNDAR  AND LEWIS 
the num be r of public ation c itations ac ross 169 unive rsities in Canada, the U .S.,
and the U .K . T hey found that a strong assoc iation existed f or individual univer-
sities in the  num ber  of its academic staff , the number of  its re se arch students,
the number of its libr ary books and journals, the level of its university re venue ,
and the number of  its institutionally associated publica tions. A  sum mar y of all
these individua l and organizational attr ibutes derived f rom  pa st r ese arch can be
seen in the profiles reported in Ta ble 1.
A n impor tant issue  to ke ep in m ind as these studies ar e review ed is the  basic
notion of ca usality. T he a bove c onc eptual f ramew ork assumes tha t individual,
depar tmental, and institutional c ha rac te ristics ar e causally prior to the average
publishing productivity le vels of  the ir  fac ulty. T here are, of c our se , notable
exce ptions. We ca n note, for  example , the many ca ses w her e distinguishe d fac -
ulty have bee n r ecr uited af ter  the ir  gr ow th at othe r institutions, w hich is espe-
cially true in the hiring practices of strong private r ese arc h universities. T here
ar e also the possible rec ursive effe cts relating to the a ssociated a nd inte rac tive
re la tionships be tw e en fac ulty size  and obtaining r esour cesÐfor  example, only
w ith a dditional resource s is a  larger f aculty cohort possible. Sever al studie s
have f ound that de partm e ntal structur es, norms, and resources (i.e., the context
in w hich fac ulty w or k and the  f acilities a nd assistants tha t ar e availa ble)  a re
im portant for  enhanced scientif ic  pr oduc tivity. L ong (1978) , f or  example, ad-
vance d the  ea rlie r w ork of C ole and C ole (1972) on a ccumulative advantage 
TA BL E 1. Attrib u tes A ssociated  w ith  R esearch  Prod u ctivity 
Individual Attributes 
Innate abilities (i.e., IQ , personality, gender, and age)
P ersonal environm ental influences (i.e., quality and culture of graduate training, and
culture of em ploying departm ent) 
Institutional and D epartm ental Attributes 
Institutional structure and leadership
S ize of program  and faculty
C ontrol by private sector
A m ount of university revenue
Availability of technology and com puting facilities
N um ber of books and journals in library
D epartm ental culture and w orking conditions
Workload policies
Availability of leaves, travel, and institutional funds for research
N um ber of students on research support
Availability of ªstarº faculty
Availability of nongovernm ental research funds 
Sources:  P ast studies on research pr oductivity.

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION615 
and found strong eff ects of departmenta l loc ation on subsequent re se arch pro-
ductivity. More  re cently, Tautkoushian, D undar, and Be cker (in pre ss) also
found that depar tmental r ese arch produc tivity w as a significant pre dictor of its
re putation and prestige , and this in tur n of te n leads to more re sourc es. 
METHOD
The Data 
D ata in this study com e from the latest 1993 N ational R esearc h Council
(N R C) study on re se arch-doctora te  programs in the U nite d States ( G oldberger
et al., 1995) . T he  original N RC  study examine d m ore  tha n 3,600 doc tora l pr o-
grams in 41 fields in the biologica l sc ie nces, the  physica l scie nces a nd m a the -
matics, the socia l a nd behavioral sc ie nc es, engine ering,  and the a rts and hum an-
ities at 274 universities. T he N RC  study based its a na lysis on many objective
and subjective measur es tha t var ie d by disc ipline  acr oss 17 to 20 m ea sure s. T he
four r eputa tional m ea sure s employed w e re (1)  the r ating of gra duate f aculty
quality, (2)  the  ra ting of gr aduate progra m  eff ective ness in tra ining resear chers
and scientists, (3) the improvement of the gra dua te program in the  last 5 ye ars,
and (4) fa m ilia rity w ith the w or k of  the program fac ulty. T he objective  mea-
sur es w er e r elated to the achievements of  f aculty in each program ( e.g., pub-
lic ations a nd citation counts) and the chara cte ristics of the pr ogr am's graduates.
In our curr ent study addressing re sea rch productivity w e attem pted to cluster
only on pr ogr ams  that had sim ilar resea rch output goals. O ur  final N RC  sam ple
included only a  tota l of 30 program  ar eas w ithin four c luster fields identified a s
the biological sciences, the  physical sciences a nd m a thematics, the social and
behavioral sciences, and the f ield of e ngineer ing f rom  acr oss the selec t 90 re-
searc h classified unive rsities in the U .S. (classified a s R esearc h I univer sities
acc ording to the Ca rnegie classif ica tion). T hese institutions all exhibit a strong
orientation tow ard r esearc h and sc holar ly w r iting and focus heavily on gra dua te
education w ith large numbers of  gr aduate pr ogr am s. M oreover, they a re typ-
ically considere d a  hom ogeneous set of institutions with similar  production
technology (B raxton a nd B ayer, 1986, p. 26). In the final analysis of  the  curre nt
study, a ll of  the field-re late d doctor al study programs in the 90 re se arch univer-
sities in the  U .S. w e re included in the study and these totaled 1,841 doc tora l
programs. We  did not include severa l f ields fr om  the a rts a nd humanities due to
their  differ ing perf orm ance indica tors.
Beyond updating pr evious studies and intr oducing new  m e asur es f or  e xplana-
tion and possible prediction, our study attem pts a new  a nd la rger  sample design
based on c luste rs of departm e ntal pr ogr ams. M ost past studies have attempted
to examine  resea rch pr oductivity by exa m ining depar tm ental prof iles, typically
in a single progra m  are a such a s politic s or e conomics. A  few  othe r r esearc h
productivity studies have also tended to f oc us on the e ntire institution. T he

616DUNDAR  AND LEWIS 
cur rent study is the  first c omprehensive  analysis of  resear ch produc tivity by
field or  cluster s of  re la te d departments a cross universities. We include d the
depar tmental programs in our  four f ields by controlling diffe rences am ong pro-
gra m s through the use of program-specific dum my variables. O ur design tec h-
nique  permitted us to examine  a larger sample of pr ogr am s. With a larger sam-
ple size , our regre ssion and statistical techniques w ere  much m ore  r obust w ith
highe r degree s of re liability. 
The Model 
Resea rch produc tivity is conventionally measured as the  r atio of tota l pub-
lications to number of pr ogr am fac ulty. Publica tion analysis of jour nal articles
and books is clea rly the most com mon measure of such resear ch per formance
(O lson, 1994). T he pr inc ipal de pe nde nt va riable f or  r esearc h productivity in the
cur rent study w as the numbe r of journal a rtic le s pe r a ve rage faculty m e m ber
attributed to ea ch of the pr ogr am s be tw e en 1988 and 1992. U nfortunately, w e
w er e only able  to include journal articles in the study since data we re not
available for books. A lthough an analysis of  journal ar ticles presents its ow n set
of pr oblems (r ela ting to suc h m a tter s a s journal quality, types of publica tion,
and m ultiple author ship), it has becom e  the m ost c ommon m e asur e of resear ch
productivity la rgely because of the ava ilability of suc h da ta  from the  Institute
for  Scientif ic  Information (B raxton a nd B ayer, 1986) .
Based on our review  of  the resear ch lite rature and a va ilability of  data, sever al
explanatory variables w ere  included in our  m odels f or examina tion. We as-
sum ed that r ese arc h productivity as a dependent var ia ble w as la rgely m ea sure d
by jour nal publications a nd that this output measure w as functionally re la ted to
those individual f aculty and organizational attr ibutes disc ussed and identif ie d in
Table 1. U nf or tunately, severa l im portant dim ensions that others ha ve  identif ied
w er e m issing f rom  our data. We are  missing appropriate m e asur es f or such
attributes as individua l fa culty innate abilitie s and their pe rsona l environm ental
influences. M oreover, w e a re missing m ea sure s that r elate to some impor ta nt
dim ensions of depar tmental culture a nd w orking conditions tha t include  w or k-
load and discr etionary de partm e ntal f unds, and the ava ilability of nongover n-
menta l r esearc h funds. I n addition, w e are  m issing da ta  on othe r institutiona l
attributes that measure the a m ount of  univer sity revenues and the availability  of
technology and com puting facilities. N evertheless, w e do have m ea sure s for an
im portant number of the re le vant departmental and institutional va riable s that
w er e uncovered in our re view  of  the literature. In severa l of  our  m odels w e
include d m ea sur es for size by controlling f or num ber s of fa culty a nd ratios of
gra duate students to faculty. We also examine d w hether the institution w a s pri-
vate or public , the amount of expenditur es for libr ary ac quisitions, the a ge  and

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION617 
seasoning of the  f aculty through per centage  of fa culty w ho w e re full pr of essors,
the scope of w hethe r many f aculty we re publishing or w hether only a  fe w
ªstarsº w er e contr ibuting to the pr ogr am 's re sea rch productivity, the perc enta ge
of fac ulty under independent re sea rch support, and the degree  to w hich gradu-
ate students w e re being suppor te d a s re se arch assistants in the progra m s. 
P rogram  size  w as mea sure d by the total num ber  of fa culty a ff iliated w ith
eac h of the  progra m s. It w a s use d mainly f or e stim a ting ave rage fac ulty produc-
tivity. 
Concentration and pe rce ntage of faculty publishing  w e re measured thr ough
the use  of the N ational Resear ch Council's G ini coefficient and a c onstructed
measur e for pe rcentage of f aculty publishing in eac h of the departm e nts. We
employed differ ent regr ession models f or  both measures. For perce ntage  of fac -
ulty publishing w e include d a ll of the progra m 's fac ulty and use d this second
measur e as a proxy for the G ini coe fficient as to w hether f aculty publica tion
distribution might be the source  of the program's resear ch productivity. It is
expected that the la rger  the  perc entage of fac ulty publishing, the la rger  the
num ber  of f aculty involve d in resear ch and schola rship ac tivity. A  pr ogr am's
re sea rch pr oduc tivity is ther efore expected to be positively re la ted to the  per-
centage of its publishing fac ulty. 
P ercentage of faculty  who are  full profe ssors  is intended to m e asur e the
degree  of fa culty maturity and exper ie nc e w ithin each of the program a reas. It
is a ssum e d that more e xper ie nced fa culty w ill be both m or e productive and
publish more (see  L e vin and Stephan, 1989). 
Institutional library  e xpenditures   w as intended to serve as a proxy for institu-
tional resources. It is expe cted that a  strong and positive relationship betw een
re sea rch productivity and libra ry expenditures exists since libra ry expenditures
ar e like ly to ha ve  a positive impa ct on the quality a nd qua ntity of such re se arch
output. 
R atio of graduate  stude nts to fac ulty  is intended as a  pr oxy for depar tmental
w orkloa d. T he number of  graduate students is assumed to influe nce the tim e
that fac ulty have to addr ess their  total w or kload. A s the  num ber  of stude nts
increa ses, resear ch time is like ly to be com e  sca rce r, and it is e xpec te d that a
high ra tio w ill result in low er re se arch perf ormance. A  highe r ratio a lso sug-
gests that f aculty may be less ac cessible to stude nts, w hich m a y r esult in r e-
duced teaching e ff ective ness. 
P ercentage of faculty w ith rese arch support  m ea sure s w hether  faculty m em-
bers are  a ctive w ith r esearc h f unded fr om outside  the depar tment. It is assumed
that the m or e active  and successful a  pr ogr am 's fac ulty is in obtaining exte r-
nally sponsore d resear ch funding, the  higher  their  re sea rch pr oductivity and
subsequent public ation rec ord. 
P ercentage  of graduate students who hold research assistantships   is a ssum ed

618DUNDAR  AND LEWIS 
to positively affec t a program's resea rch pe rfor m ance . T his pr oposition is al-
leged to be  particula rly tr ue  in the science a nd e ngineer ing  fields. Such stude nts
often help f aculty inc rea se their resear ch pe rformance  by dir ectly contributing
to pr oje ct-related re se arch. 
Institutional control on rese arch pe rform ance  is dir ectly examine d thr ough
the use of  a dumm y va riable  that indicates w hether  the  institution is a  public
unive rsity or a private sc hool. 
The General Statistical Model 
T he genera l mode l f or  this study ca n be e xpr essed as follow s: 
P 
i 
4 a 
0 
` a 
1 
F 
i 
` a 
2 
F 
i 
2 

` 

7 

( 

j 
4 
3 

a 
j 
X 
ji 
` e 
i 

w her e  P 
i 
 is the num ber  of aver age resear ch publica tions of eac h departmenta l
cluste r at institution  i ;  F 
i 
 is the num ber  of depar tmental f aculty clusters at insti-
tution  i ;  X 
ji 
 is the number of other  expla na tory va riable s a ssum ed to influe nce
depar tmental resear ch pr oduc tivity;  e 
i 
 is a ra ndom e rror  ter m ; and  a 
j 
 is the
coefficient to be estimated.
T he nature of  the  individual e xplanatory  var ia bles  is derived fr om  our r evie w
of past litera tur e a nd is identified in Tables 1 and 2. T he specif ic form in w hich
these influencing f actors is m ea sure d is taken or constructed fr om the  data sets
available to the study.
We  can test w hether the r elationship  betw een a ve rage publication a nd depa rt-
menta l size is nonline ar by examining the squared fa culty siz e variable. H old-
ing all other  variables constant, if  a 
1 
 proves to be positive  and  a 
2 
 pr ove s to be
negative, the avera ge  de pa rtm ental publication w ill first incr ease as f aculty size
incre ase s, but at a diminishing rate.
We also should note that depa rtm ent-specific dumm y var ia ble s a re included
in all the mode ls to c ontrol for  diffe rences am ong the depar tm ents. T he m odel
specified above w as fir st estimated for all the depar tm ents com bined and the n
run for  each of the four f ie lds (i.e., biology, enginee ring, physic al sc ienc es, and
social and behavioral science s) . T hese five mode ls a re identified in Table 3.
S everal other mode ls based on this gener al m odel w ere  a lso estimated in the
study and are  identif ie d in the  follow ing disc ussions.
It should be noted that w e also atte m pted to examine the da ta  w ith re spec t to
ea ch of  the 30 individual depar tmental pr ogr am s. H ow ever, the data appea red
not to fit our models very w ell and resulted in very low   R 

2 

  a nd in several of the
ca se s the  mode l equations w e re not signif ic ant. T his c an be attributed to the
sm a ll sample size s in most of  the  departmenta l program ar eas. A s a conse -

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION619 
quence, w e  used the cluste r fields as our mode l of analysis by controlling f or
departmental diff ere nc es with individual dumm y var ia bles. 
RESULTS 
By employing all of  the progra m  ar eas availa ble in the  f our  cluster f ie lds, our
final sample  included 1,834 programs in the N RC data set. P rograms included
in the pooled sa m ple w ere  534 in the  biological sc ie nc es, 380 in e ngineer ing,
481 in the physica l sciences and mathematics, a nd 438 in the social a nd behav-
ioral sc ie nc es. Fr om  data supporting Ta ble 2, w e note  that aver age program
fa culty siz e var ie d from 24 in e ngineer ing to 36 in the biologica l science s. We
also note that there  w er e consider able diffe rences be tw e en rates of  produc tivity
betw een fields as measured by the aver age public ations of eac h program
w here in they var ie d from 2.5 in the social and behavioral science s to 9 in the
biologic al science s. N ot surprisingly, a typica l f aculty membe r in the soc ia l
sciences w rote 2.5 ar ticles betw een 1988 a nd 1991, w her eas a colleague in the
biologic al sciences w r ote about 9 ar ticles during the same per iod. T hese diff er-
ence s m a y not r eflect produc tivity differ ences betw e en f ie lds but m ay simply
indica te  differe nces in styles and types of  publications be tw e en fields.
E m ploying our gener al mode l for  e stim ating influences a nd assoc ia ted fa c-
tors w ith de pa rtm e ntal resea rch productivity, w e fr amed five  sepa rate re gr es-
sions. I n Table  3 the se  five estimated re gressions are pr esente d. We f ir st esti-
mate d results f or  the entire sample  combined a nd then subse quently estimated
separa te mode ls f or each of the four cluster  fields of re sea rch. E ac h of the
models also includes department- spe cif ic dumm y var ia bles, but these ar e not
re por te d in Table 3 due to the ir  length and lim ited importance to our questions.
O nly 6 ( i.e., molecular and gener al gene tics, neuroscie nces, biomedic al engi-
neer ing, astr ophysic s a nd astr onomy, physic s, and sta tistics a nd biostatistics) of
the 30 departments w ere  f ound not to be significant in the  f ull m odel estim a -
tion. T he estimated and adjusted  R 

2 

 for the pooled m odel w as a relatively high
.67 a nd for the f our  fields the   R 

2 

 ranged f rom  .50 for the  biologica l sc ie nc es to
.80 f or the social and behavioral sciences, all indicating reasonably good f its
betw een our models a nd the data set. It is im portant to note that all of the
coefficients ( except for the six depar tmental dumm y variables ide ntified above)
for the f ull m odel are  statistically signif ica nt at ver y high le ve ls w ith all asso-
ciations in the  expec te d directions, indic ating ve ry high r elia bility for our esti-
mate s. Both the coefficient signs and degree s of significance appear to be con-
sistent ac ross the four cluster  fields, only w ith som e mate rial diff ere nc es in the
soc ia l science f ie lds.
With these mode ls conceptually fr amed from our  review  of the past lite ra-
ture, w e should now  be better  a ble to address som e of the persisting policy

620DUNDAR  AND LEWIS 
TABLE 2. Descriptive Statistics All FieldsBiologicalSciencesEngineeringPhysicalSciences andMathematicsSocial andBehavioralSciencesMeanMeanMeanMeanMean(Std. Deviation)(Std. Deviation)(Std. Deviation)(Std. Deviation)(Std. Deviation)Average faculty publications6.659.037.147.322.57(4.76)(4.12)(5.18)(4.64)(1.80)Faculty29.8536.5824.2728.3228.07(21.82)(31.19)(15.53)(16.94)(13.72)Faculty-squared1366.752308.89829.261088.80975.50(2902.50)(4818.92)(1236.12)(1564.34)(1082.28)Ratio of graduate students to faculty2.371.293.012.722.77(1.95)(1.18)(2.36)(2.26)(1.29)Percentage of graduate students who are30.5726.5350.6837.8010.23RAs(26.18)(22.86)(21.94)(27.44)(12.22)Library expenditures ($, 000)15961.6615500.0015607.3816228.2016540.24(8190.55)(8196.29)(7157.63)(8830.09)(8272.81)Public institutions (1 if public, 0 otherwise)0.700.690.730.690.70(0.46)(0.46)(0.45)(0.46)(0.46)Percentage of full professors56.0952.6756.0060.9155.10(15.24)(14.98)(14.88)(16.13)(13.45)Percentage of faculty publishing78.8086.9979.2782.4364.31(15.52)(9.13)(12.70)(11.53)(17.78)Percentage of faculty with research support49.8664.4853.2858.6719.24(26.07)(20.24)(19.89)(19.50)(17.08)Valid  N  (listwise)1834534380481438 Source:  Goldberger et al. (1995).

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION621 
TABLE 3. Estimates of Regression Coefficients Explaining Average Faculty Publications VariablesFull ModelBiologicalSciencesEngineeringPhysical Sciencesand MathematicsSocial and BehavioralSciencesCoefficientsCoefficientsCoefficientsCoefficientsCoefficients( t -statistics)( t -statistics)( t -statistics)( t -statistics)( t -statistics)Constant 1 4.29 1 10.63 1 9.20 1 3.930.19 1 (6.76)* 1 (7.78)* 1 (5.72)* 1 (3.55)*(.54)Faculty0.050.060.0480.070.01 1 (5.99)*(4.70)*(1.21)(3.07)*(1.17)Faculty-squared 1 0.0002 1 0.0003 1 0.0003 1 0.001 1 0.0001 1 (3.50)* 1 (3.51)* 1 (.63) 1 (2.27)* 1 (.65)Ratio of graduate students to faculty0.160.190.170.26 1 0.05(4.20)*(1.59)(2.12)**(4.42)* 1 (1.82)**Percentage graduate students who are RAs0.0090.0020.0280.010.01(2.74)*(.35)(3.156)*(2.00)*(1.30)Library expenditures0.000040.00010.0000010.00010.00001(4.84)*(5.76)*(.14)(3.84)*(.88)Public Institution (1 if public, 0 otherwise) 1 0.79 1 1.620.04 1 0.65 1 0.15 1 (5.37)* 1 (5.56)*(.094) 1 (2.39)* 1 (1.57)Percentage of full professors0.0290.050.060.02 1 0.0005(6.07)*(5.19)*(4.22)*(2.41)* 1 (.151)Percentage of faculty publishing0.080.150.070.070.04(11.77)*(9.47)*(3.77)*(5.46)*(12.64)*Percentage of faculty with research support0.040.030.070.030.03(10.26)*(4.49)*(5.98)*(4.18)*(8.90)*Adjusted  R 
2 
0.680.510.570.710.79 F -statistic102.7937.7932.675.12110.72Sample size1,834534380481438 Note:  All models include additional program specific dummy variables.Tests of significance: * p   ,  0.01, ** p   ,  0.05.

622DUNDAR  AND LEWIS 
questions in higher education. O ur  results should be espe cially helpf ul w ith
re gard to depar tmental a nd institutional policy on how  such institutions might
enhance  their r ese arc h productivity. 
Is a larger  progr am w ith more facu lt y more  produc tive?  T he poole d
mode l inc luded all aca dem ic programs fr om  all four of our  diff erent field clus-
ters: the social and be ha viora l sc ie nces, the physical sciences a nd m a thematics,
engineering, and the biological sc ienc es. T he re sults indicate that pr ogr am and
depar tmental size in numbers of fac ulty is, indeed, a  strong pr edic tor  of  average
depar tmental publication. H ow ever, the  negative sign on our f aculty- square d
coefficient  acr oss all five of our models indicates tha t the effect of progra m  siz e
dim inishes a s siz e incr eases. With such r esults it is also logic al to a ssum e  that
beyond some large f aculty size, suc h avera ge productivity w ould not only f ail
to r ise but w ould ac tually dec line. U sef ul ref ere nc es to faculty sizes a nd their
sta nda rd de viations acr oss all the  f ie ld cluster s ca n be gained f rom  data in Table
2. 
D oes the  type  of institu tional cont rol make  a diff ere nce?   We estim a te d for
the effe ct of type  of institutiona l contr ol ( i.e., public or  priva te ) by including a
dum my var ia ble in ea ch of  the models. We  found in all our estimated r egression
mode ls tha t if the  institution wa s a public unive rsity this type of contr ol assoc i-
ated negatively w ith depa rtm ental r esearc h production. In othe r w or ds, aver age
fa culty r ese arch pr oductivity in public univer sitie s tends to be  signif ic antly less
than in private institutions, contr olling f or  faculty size and the sever al other
var ia bles.
T here  m ight be severa l plausible expla na tions for w hy pr ivate institutions
appea r to have gre ater re se arch productivity on the part of the ir  faculty. F ir st,
this m a y be due to fac ulty in public institutions producing other products (e .g.,
producing othe r types of non- journa l-related r esearc h products or f ocusing on
teac hing or  mor e outrea ch r athe r than resear ch). Mor eover, as note d by Jor dan
and his collea gues (1989) , public universities m a y be politica lly limited in the ir
attempts to increa se their re sea rch productivity beca use of their public natur e
and goals. Sec ond, ince ntives for private r ese arc h universities  to maxim ize their
re se arch per formances and reputations of ten are ver y strong because most pri-
vate schools ar e heavily de pende nt on voluntar y suppor t and de m a nd f or  the ir
products in or der to gene rate reve nue. Since high re sea rch perf orm ance often
re sults in significa nt re putational increa se  a nd thus additional r esour ces, most
private institutions attempt to re cruit the m ost re sea rch-pr oductive faculty. Jor-
dan and his colle agues (1989) desc ribed this stra te gy as a ªsurvival testº f or
private unive rsities. Most pr ivate r esearc h universities genera lly have few e r but
more highly resea rch-productive  fac ulty than those typically f ound in public
schools. F ina lly, it is also like ly that m any private institutions may provide
better organiz ational str uc tur e and perf or m anc e inc entives for  faculty to en-

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION623 
hance their r ese arc h productivity. We do know  that privates at lea st provide
higher  sa la ries ( Clotf elter et al., 1991). 
D oe s havin g more fu ll prof essors mak e a d ep ar tmen t more prod uct ive in
re se arch  p erforman ce?   T he eff ect of  f aculty rank on resear ch pr oductivity w a s
found to have m ixed results in pa st liter atur e: Seve ral studies f ound that rank
w as a signif ic ant pr edic tor  of f aculty r esearc h pr oduc tivity a nd sever al other
studies found that rank had no e ff ect (see Tie n and B la ckbur n, 1996, p. 3).
A lthough w e  do not ha ve  a  va riable m e asur ing alternative forms of f aculty
ra nk, w e do ha ve  a m e asur e tha t indic ates the per centage of depar tmental fac-
ulty w ho w ere  full prof essors. We  a ssum e that full pr ofessor s a re tenured, expe-
rience d, and matur e senior  fac ulty. If a  program has a large  perce ntage of full
profe ssors, w e c an e stim ate the likely effec t of this ra nk and seniority of pr o-
gram f aculty. We  found that having a higher  per centage of  f aculty m embers
w ho a re full prof essors indeed does help to achieve higher  resear ch pr oductivity
in almost all f ie lds. E xcept in the social a nd behavior al sciences, all the e sti-
mate d c oeff ic ie nts are  positive and statistically significa nt. 
W h at is the  ef fec t of  ªstarº facu lt y on rese arch  p rod uct ivity?  S om e past
re sea rch has f ound that the  resear ch productivity of a de pa rtm e nt c an be influ-
ence d consider ably by the pre senc e or  pr oductivity of  ªstar º fac ulty ( Johnes,
1988a ). Cole and Cole  ( 1972) have re por te d, for example, that the  most inf luen-
tia l resea rch being produced in m any fields is being conducted only by a small
num ber  of all those engaged in resear ch activity. We  used a N a tional Resear ch
Council constr uc te d G ini c oeff ic ie nt for public ations to test for the  possible
effec t of ªstar º fac ulty on depar tm ental avera ge resear ch productivity. T he G ini
coefficient  he re is used to m e asur e the degr ee of publication conce ntra tion on a
single or small number of progra m  f aculty dur ing the per iod 1988±1992. A s
G oldberger a nd his colleagues (1995, p. 56)  ha ve  noted: 
The higher the G ini coefficient, the less the dispersion. F or exam ple, consider three
program s each having 20 faculty m em bers w ho together publish 40 articles during a
certain period of time. In program  A , each of the 20 faculty m em bers published 2
articles for a G ini coefficient of 5. In program  B , 2  faculty m em bers each produced 11
articles . . .  and the remaining 18 faculty m embers each published one article; the G ini
coefficient w ould be 16.3. In program  C , one faculty m em ber published 40 articles
and the remainder published none; the G ini coefficient w ould be 100. 
Such a measur e c an be used to te st for  the  inf luence  of ªstarº fac ulty in deter-
mining de pa rtm e ntal r ese arch pr oductivity. If a sm all num ber  of  fa culty mem -
bers ar e highly productive  and the depar tment a lso appear s to be  highly produc-
tive, the n the re cruitment of  ªsta rº fa culty appear s to be  a rationa l and use ful
policy.
Fortunately, the  1995 N R C da ta  set r eported on suc h constr uc te d G ini coeffi-
cients f or  each of the departmenta l progra m s in the total sam ple. T his m ea sure

624DUNDAR  AND LEWIS 
per m itted us to te st for  this ªsta rº f aculty e ff ect acr oss the  leading r esearc h
unive rsities in the U nited S ta te s. W hen this G ini coefficient w as substitute d for
our measure of perc entage  of fac ulty w ho w er e publishing w e  f ound that it only
w as statistically significa nt and ne ga tive in our pooled and physic al science s
and mathe m atics estimate d m odels. T hese  results indicate that ve ry few  univer-
sity depar tments a re r elying on only a fe w  ªstar sº to c arry their effor ts in r e-
sear ch pr oduc tivity. 
W h at is th e e ffe ct of  d istrib ution of  fac ulty act ivity on  re se arch  produc -
tivity?  It is assumed that a large per centage of  fac ulty publishing in a de pa rt-
ment is likely to ha ve  a statistically signif ic ant and positive influence  on the ir
depar tmental resear ch productivity. N ot surpr isingly, this factor w a s f ound to be
both highly significa nt and positive  in a ll our  mode ls. A  de pa rtm e nt that w ants
to incr ease suc h pr oductivity ought to expect contribution f rom  a ll of its m e m -
ber s ra ther  than f rom  only a f ew  ªprolif icº or ªstarº f aculty. 
W h at is t he ef fect  of facu lt y w ith  fin an cial re search supp ort?  We also
examined for the  effect of  fac ulty w ith nondepar tmental financial re sea rch sup-
port on depar tmental re se arch pr oduc tivity. We assum e d tha t the  higher  the
per centage of fa culty w ith such financial support, the  higher  w ould be  the ir
re se arch per formance. N ot surpr isingly, we  found that this fa ctor  is closely re-
lated to depar tmental pr oduc tivity in a ll four of the cluster  f ields. 
W h at is th e eff ect  of  in st it utional su pp or t on critical resources?  Unf or-
tunate ly, the only m ea sure  available f or us to e xam ine this question w as institu-
tional expenditure for  their campus librar ie s. We found tha t in all c ases, e xc ept
engineering and the social sc ie nc es, institutional expenditures f or  libra ries w e re
significantly r elated to depar tmental resear ch productivity. Be yond the  indirec t
support of  libr arial holdings in support of de partm e ntal r esearc h, this positive
effe ct also m ay suggest tha t institutions w ith more resources pr ovide be tter
re sourc es in m a ny other infra structure w a ys as w ell. Most, if  not all, of this
infra structure support should contribute to incre asing their resear ch produc-
tivity. 
D oes fac ulty t eachin g load  w ith  gr ad uate  stu den ts inf lu en ce research
p rodu ctivity?  T he  N RC data set permitted us to construc t a  mea sure  for the
ra tio of gr adua te  students to fa culty and to examine w he ther  suc h a high r atio
might r educe average  re se arch produc tivity. O ne likely sce na rio re sulting f rom
a high r atio is that f aculty mem be rs may have le ss time to do resear ch since
they m ight be r equir ed to teac h m ore c our se s or a dvise m ore  students. O n the
other hand, an alternative possible sc enario re sulting fr om a high ra tio is that
fa culty and stude nts c an c ollaborate in projects a nd conduct joint resear ch a nd
subsequently publish their r esults. L a rge numbe rs of gr aduate students also
might contribute  to a richer m ix of sc holar ship that might re sult fr om labora to-
ries, se m inars, and thesis pr oje cts. T his may be par ticularly impor tant in the
science-  a nd enginee ring-related fields as w as noted by L odahl and G or don

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION625 
(1972), w herein they found that high par adigm f ields such as e ngineer ing a nd
physical sc ie nces do in f act use gra duate students more effe ctively in both their
teaching and r esearc h a ctivities.
In our analysis of depar tmental re se arch pr oduc tivity, w e found tha t a high
ra tio of  gra dua te students to f aculty is statistically and positively associate d
w ith depar tmental resear ch produc tivity in the pooled m odel as w ell as in engi-
neer ing a nd the physica l sc ienc es. Somew hat surprisingly, in the socia l and
behavioral sciences w e f ound that ther e w a s a significa ntly negative a ssociation
betw een the ratio of gr adua te  students to fac ulty and depar tmental re se arch
productivity. T his la tter  effe ct undoubtedly re sults from w orkload eff ects and
the prospe ctive problems  of com munica tion for low -para digm  fields such as the
soc ia l sciences (L odahl and G or don, 1972). I t a lso indicates the  differ ential
effec ts tha t occ ur  across the diff ering f ie lds and clusters of departmenta l ar eas. 
W h at is the  eff ect  of employing gr ad uate  st ud ent s as rese arch  assist an ts? 
A cr oss all of our models w e found that the per centage  of graduate stude nts w ho
held resear ch assistantships w as also positively associated w ith de partm e ntal
re sea rch productivity. T he  estim a te d c oeff ic ie nts all ha d positive signs and w ere
sta tistica lly signif ic ant in the  pooled as w e ll as in the engine ering a nd physical
sciences m odels.
It is important to note that w e  a lso ran all of  the five m odels identified in
Table 3 w ith resea rch c itations per  aver age f aculty member a s the de pende nt
variable in pla ce of  journal publica tions. I n no ca se  did any of  our results
change in e ither direction of signs or  in signif icance of coefficients. T he  only
mate rial c ha nge  w a s in the low er   R 

2 

  a nd degre es of significance found w ith the
citations per f aculty. I n short, journal public ation proved to be a better  fit for
our m odels and f or our  que stions. 
Limitations 
T he findings of this study provide som e  insight on r ese arc h productivity in
A m e rican higher  education. Ye t there  a re severa l c aveats that need to be borne
in m ind in inter preting the re sults. F irst, it must be remembered that pr oduc -
tivity in higher education is a m ultidim ensiona l conc ept. W hile re se arch pr o-
duction is an important dim e nsion of productivity, it is only one of se ve ral
major  outputs (e.g., resea rch along w ith teaching, public ser vic e, and outre ach)
in m ost resear ch univer sitie s and some of these other outputs might have highe r
soc ia l or  politic al priorities at tim e s. A n inc rease in the teac hing load of a
department, for exa m ple, is likely to le ad to re duc ed resear ch pe rfor m ance  due
to tim e  constraints.
Sec ond, e ve n though r ese arch often appear s to be the e asiest of  higher e duc a-
tion's output to measure, the re are  still some ser ious difficulties in measuring
re sea rch perf or m anc e. E valuating resear ch per formance is an ine xa ct science

626DUNDAR  AND LEWIS 
sinc e the re is no prec ise measure of  r esear ch output. T here a re multiple f orm s
of resear ch outputs such as journal a rtic le s, books, book chapte rs, monogr aphs,
unpublished confe rence pr ese ntations, a nd even computer  softw ar e. T his study
w as only able  to use I nstitute of  Scientific Inf or m a tion [I SI]  jour na l article s as
its main measure of re se arch output since data related to the othe r forms w er e
not readily availa ble. O lson (1994) has note d, for  example, tha t the f aculty in a
par ticular unive rsity can often appear  m ore  pr oductive tha n f aculty in a nother
institution w hen only one product is analyzed, and this per ceived re sult may be
misleading due to the existence of an alternative f orm  or measure of resea rch
productivity. T he significance of ar ticle publica tion r ates should not be exagger-
ated.
It a lso must be  clea rly acknow ledged that the c ur rent study used only journal
ar ticle counts tha t we re obtained fr om  the  Science  Citation Index   a nd the  Social
Science  Citation Index  as its measure of output a nd articles. T his sourc e of data
may have resulted in sever al biase s. T her e ar e sever al proble m s r elating to the
use of public ation and citation c ounts from these  types of indexes ( see  Johnes,
1988b) . I t is im portant to note that journal articles not c ove red by those indexes
w er e not included. Fur ther m ore , no control w as m ade  f or  quality since  no m ea -
sure w as available  f or suc h purposes. T he IS I m a intains a computer  file  consist-
ing of  bibliogra phic re cords of  pa pe rs in the  tw o citation inde xe s. From 1988
through 1992, the IS I matched its citation counts w ith institutional faculty lists
fr om  1992 that w e re provide d by the  N a tional Resear ch C ounc il. T hus, if  fac-
ulty members m oved during the  5-year  pe riod, they w ould only be counted in
the re ceiving depar tment and the n only w ith c itations fr om  their tenure within
the rece iving depar tment. T his w ould not be a pr oblem if  fa culty rar ely changed
depar tments, but acade m ics in resear ch depar tments are a mobile gr oup in the
labor forc e. T hree  types of departments are  disa dva ntaged by these da ta -gather-
ing pr ocedures: those tha t e xper ie nced grea ter  tha n a ve rage e m igration; those
that e xpe rienced grea te r tha n aver age im migr ation; and those that hir ed more
new  P h.D .s than aver age beca use they w ould be adding to their sta ff  w ith larger
num ber s of  fac ulty w ith lim ited produc tivity. O n the othe r hand, those  depart-
ments tha t w ould benefit in articles per fa culty ratio comparisons m ost likely
w ould be  in distinguished institutions w ith older  professor s (large propor tions
of f ull profe ssors, w ho are  le ss likely to m ove). It is c lea r that this pr ospec tive
bias ne eds f ur the r examina tion. N ever theless, although the  use of journal pub-
lications as the sole  indica tor of depar tmental resea rch productivity can be  cr it-
icized, such a measure f or  publication is an impor tant a ve nue  for disseminating
re se arch re sults and this measure has strong a cceptance ac ross m ost a cademic
peer  gr oups. Suc h measures are also used for aca de m ic appointm e nts, promo-
tions, and othe r rew a rds.
T hir d, the data available  f or  this study lim it the development of  a more com -
pre he nsive r ese arc h produc tivity m odel. C ross- sec tional data a nd designs often

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION627 
provide  misleading results w hen c ompare d w ith the em ploym ent of a more
appropriate longitudinal design ( L ong, 1978). Mor eover, there are seve ral othe r
measur es related to individua l faculty and departmenta l and institutional attri-
bute s tha t ar e likely to have an influence  on a nd contribute to the r esearc h
perf orm ance of  depar tmental programs. N ot only does som e suggestive past
liter atur e a nd our  ow n intuition sugge st that the intelligence, aptitude, exper i-
ence , and enthusiasm  of the aca de m ic staff  themselve s should have some influ-
ence  on resear ch produc tivity but suc h r elia ble  attr ibutes are  very diff icult to
measur e a nd are  m issing fr om  the curre nt study. Moreover, the  leader ship of a
department a nd its orga niza tional and adm inistrative effectiveness ca nnot be
easily qua ntified and measured. D iff erences in pr oductivity levels m a y ref le ct,
for example, var iations in tea ching loads and the  availability of other r esour ces
for r esearc h.
Finally, the re also is evide nc e of the  existence of ec onomie s of scope in
higher  education that r esult from the  joint production of re se arch and instruc-
tional activities, espe cially w ith regar d to gr aduate education in re se arch univer-
sities. A ny analysis of resear ch productivity w ithout c onside ring the  additional
effec ts of  r esear ch production on teac hing and lear ning is likely to under esti-
mate  the  full eff ect of such re sea rch ac tivity. T he r esults of  this study, along
w ith all othe r sim ilar w ork, need to be view e d w ith an a ppr opr ia te amount of
caution due to these lim ita tions. 
DISCUSSION AND POLICY RECOMMENDATIONS 
In this paper  w e examined r ecent da ta  f rom  the N R C study on A m er ic an
doctor ate-level re se arch progra m s in order  to investigate the relationships be-
tw een de partm e ntal resea rch pr oduc tivity and institutiona l fa ctor s and policies.
T he f indings from this study a re important for U .S. re se arch universities f or
severa l r easons. Most impor ta ntly, w e found consiste nt w ith some pre vious
studies that aca dem ic resear ch productivity is close ly associated w ith program
fa culty size , but usua lly at a  dim inishing  r ate. P rograms w ill achieve incre asing
re sea rch perfor m ance  a s the y increa se the numbe r of  their program f aculty, but
at a  ce rtain level the m argina l product of an a dditional f aculty m e m ber  w ill
begin to de cline. Pre vious studies noted that depa rtm ental fac ulty size affec ts
re sea rch pr oductivity to som e extent as a result of  cr itical mass. In or der to
have a highly pr oductive resear ch pr ogr am , a cademic de pa rtm e nts a lso ne ed to
have a significant num ber of  f aculty. T his effe ct suggests that fa culty size  m ay
have an influence on individua l produc tivity resulting fr om  enhanced oppor-
tunity f or  c ollaboration and reinforce ment. Pr ogr ams atte m pting to inc rea se
their  resear ch pr oduc tivity (and the reby the ir  r eputa tion) should e xa m ine the
marginal pr oduct of an additional fac ulty in the ir  de te rmination of a n optimum
program size .

628DUNDAR  AND LEWIS 
T he study w as a lso useful in identifying other  f actors a ffe cting resear ch pro-
ductivity. We  found, not surpr isingly, that de partm e nts loca te d in pr ivate univer-
sities genera lly have higher  resear ch produc tivity. Public  resear ch univer sitie s
need to ide ntify those f actors affe cting this differ ential in the ir  re se arch produc -
tivity, e spec ia lly those inhibiting pr oductivity fac tors in public  r esearc h univer-
sities. A ttention ne eds to be given in the public  institutions to r em oving the se
inhibiting factors and e nha nc ing favora ble policies. We also uncovered unde r-
sta ndings tha t ha ving more full pr of essors a nd la rger  perc enta ges of depar tmen-
tal f aculty w orking on re se arch, and ha ving m ore ªstarº  fa culty, a ll contr ibute in
material w ays to enhancing de partm e ntal r esearc h produc tivity. P olicies tha t
hire mor e senior a nd ªstarº f aculty and policies that induce more existing f ac-
ulty to contribute to r esearc h w ill undoubtedly contr ibute to enhanced depa rt-
menta l pr oductivity and subsequent reputations. D e pa rtm ental and institutiona l
policies that (1) induce faculty to solic it more gr ants and c ontrac ts outside of
the unive rsity, (2) ta rget institutional resources on r esearc h produc tion suc h as
libr ary re sourc es, technology, and gr aduate student resear ch support, and (3)
re cruit a  critic al mass of  gr aduate students w ill all enha nce re sea rch produc-
tivity.
A dditional studies are  ne eded tha t exa m ine the  m ultiproduct na tur e of higher
educa tion a nd the  interac tive effects that result from  the resea rch, outr eac h, a nd
teac hing activities that frequently take place w ithin m a ny of our top re se arch
unive rsities. S uc h inf or m a tion w ould be pa rtic ularly he lpful in our understa nd-
ing of how  differ ent dimensions of  output struc ture s a nd pr oductivity change
w ith r espe ct to diff erent se ts of inputs. A  particularly usef ul a ttempt w ould be
the identif ica tion of a  full set of  outputs a nd inputs in a r elatively hom ogeneous
set of  a cademic  de pa rtm ents to be tter  under stand pr oductivity in these  depa rt-
ments. U nfortunately, the de partm e ntal da ta  sets in the N RC da ta  w er e too
sm a ll to per m it this type of study.
Pr oductivity in A merican highe r education, despite m a ny attempts to talk
about it a nd a fe w  atte m pts to measure it, is still a r elatively uninvestiga ted
ar ea. L ittle is know n about the f actors affec ting r ese arc h produc tivity, pa rtic -
ularly w ith r espe ct to its inter active effec ts w ith tea ching a nd lea rning. W hat,
for  example , is the likely contribution of  graduate stude nts to resear ch produc-
tivity? Re se arch productivity in pa rtic ular  needs a more c omprehensive  ap-
proac h to identif y factors affe cting not only individual f aculty but a depar t-
ment's pr oductivity as a  w hole . H ow  individual fac tor s a nd departm e nta l
fa ctors intera ct to r esult in a  pr oduc tive re se arch e nvironm ent is still not w ell
understood. S tudies exa m ining, for  e xa m ple, the r elationships betw ee n aging
and r ese arch pr oductivity ar e w eak a nd in need of  further investigation ( Clark
and L e w is, 1985). T his latter que stion is par ticularly impor ta nt given the ex-
pected graying effec t of A m e rican higher education over the  next de cade. In
shor t, f uture  studies examining r esearc h productivity in our  r ese arc h univer-

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION629 
sities m ust inc lude better m ea sur es for specific individual, depa rtm ental, a nd
institutional fa ctor s as the y e xa m ine policy remedies f or enhanced effec tive-
ness. 
REFERENCES 
B aird, L. L. (1986). W hat characterizes a productive research departm ent?  Research in
H igher Education  25(3), 211±225.
B aird, L. L. (1991). P ublication productivity in doctoral research departm ents: Inter-
disciplinary and intradisciplinary factors.  Research in H igher Education  32(3): 303±
318.
B ell, J. G ., and S eater, J. J. (1980). P ublishing performance: D epartm ental and individ-
ual.  Econom ic Inquiry  16(4): 599±615.
B entley, R., and B lackburn, R . (1990). Changes in academ ic research perform ance over
tim e: A  study of institutional accumulative advantage.  Research in H igher Education 
31(4): 327±345.
B lackburn, R. T., B ehymer, C . E ., and H all, D . E . (1978). Research note: Correlates of
faculty publications.  Sociology of E ducation  51(2): 132±141.
B raxton, J. M . and Bayer, A . E . (1986). A ssessing faculty scholarly perform ance. In
J. W. C resw ell (ed.),  Measuring F aculty R esearch P erform ance . N ew  D irections for
Institutional Research 50. San F rancisco: Jossey-Bass.
C artter, A . M . (1966).  A n A ssessm ent of Quality in G raduate E ducation . Washington,
D C : A m erican C ouncil on Education.
C lark, S . M ., and L ew is, D . R . (eds.) (1985).  Faculty Vitality and Institutional Produc-
tivity: C ritical P erspectives for H igher E ducation . N ew  York: Teachers College P ress.
C lotfelter, C . T.,  Ehrenberg, R . G ., G etz, M., and S iegfried, J.  J. (1991).  Econom ic Chal-
lenges in H igher E ducation . C hicago: U niversity of C hicago P ress.
C ohn, E ., R hine, S . L . W., and S antos, M . C. (1989). Institutions of higher education as
multiproduct firm s: E conom ies of scale and scope.  Review of Econom ics and Statistics 
71(2): 284±290.
C ole, J. R., and Cole, S . (1972). The O rtega H ypothesis.  Science   178 (O ctober 27): 368 ±
375.
C onrad, C . F., and Blackburn, R. (1985). Correlates of departmental quality in regional
colleges and universities.  Am erican Educational R esearch Journal  22(2): 279±295.
C onrad, C . F., and B lackburn, R. T. (1986). P rogram  quality in higher education: A
review and critique of literature and research. In J. C . Sm art (ed.),  H igher E ducation:
H andbook of T heory and Research  (pp. 283±308). N ew  York: A gathon P ress.
C resw ell, J. W. (1985).  F aculty R esearch P erform ance: L essons from  the Sciences and
the Social Sciences . A S H E -ERIC  H igher Education Report N o. 4. Washington, D C :
G eorge Washington U niversity.
C resw ell, J. W. (1986).  Measuring Faculty Research P erform ance . N ew  D irections for
Institutional Research 50. San F rancisco: Jossey-Bass.
C rew e, I. (1988). R eputation, research and reality: T he publication records of U .K . de-
partm ents of politics, 1978±1984.  Scientometrics  14(3 ±4): 235±250.
D rew, D. E ., and K arpf, R . (1981). Ranking academ ic departm ents: E mpirical findings
and a theoretical perspective.  Research in H igher Education  14(4): 305±320.
E hrenberg, R . G ., and H urst, P. J. (1996). T he 1995 N R C ratings of doctoral program s: A
hedonic model.  C hange  (M ay/June): 46 ±50.

630DUNDAR  AND LEWIS 
F airw eather, J. S . (1988). R eputational quality of academic program s: T he institutional
halo.  R esearch in H igher E ducation  28(4): 345±356.
G oldberger, M . L ., Maher, B . A ., and F lattau, P. E . (1995).  R esearch-D octorate Pro-
gram s in the U nited States: C ontinuity and Change . Washington, D C : N ational A cad-
em y P ress.
G olden, J., and Carstensen, F. V. (1992a). A cadem ic research productivity, departm ent
size and organization: Further results, rejoinder.  E conom ics of Education R eview 
11(2): 169±171.
G olden, J., and Carstensen, F. V. (1992b). A cademic research productivity, departm ent
size and organization: F urther results, com ment.  E conom ics of E ducation R eview ,
11(2): 153±160.
G olden, J., C arstensen, F. V., Weiner, P., and K ane, S . (1986). Publication perform ance
of fifty top econom ic departments: A  per capita analysis.  Econom ics of Education
R eview  5(1): 83±86.
G raves, P.  E.,  M archand, J.  R ., and T hompson, R . (1982). Econom ics departm ental rank-
ings: R esearch incentives, constraints and efficiency.  American Econom ic Review 
72(5): 1131±1141.
G runig, S.  D . (1997). R esearch, reputation, and resources: T he effect of research activity
on perceptions of undergraduate education and institutional resource acquisition.  Jour-
nal of H igher E ducation  68(1): 17±52.
H agstrom  W. O . (1971). Inputs, outputs, and the prestige of university science depart-
m ents.  Sociology of E ducation  44(4): 375±397.
H ughes, R . M . (1925).  A  Study of the G raduate Schools of A m erica .  O xford, O H : M iam i
U niversity P ress.
John, E. P  St. (1994).  P rices, P roductivity, and Investm ent: A ssessing Financial Strate-
gies in H igher E ducation . A S H E -ERIC  H igher Education Report N o. 3. Washington,
D C: G eorge Washington U niversity and A S H E.
Johnes, G . (1988a). R esearch performance indications in the university sector.  H igher
E ducation Q uarterly  42(1): 55 ±71.
Johnes, G. (1988b). D eterminants of research output in econom ic departments in B ritish
universities.  R search Policy  17(3): 171±178.
Johnson, R ., et al. (1995). S ize versus performance in research.  A ustralian U niversities
R eview  38(2): 60±64.
Jones, L. V., L indzey, G ., and C oggeshall, P. E . (1982).  A n A ssessm ent of Research-
D octorate P rogram s in the U nited States  (5 volum es). Washington, D C : National
A cadem y P ress.
Jordan, J. M ., M eador, M ., and Walters, S . J.  K . (1988). Effects of departmental size and
organization on the research productivity of academic econom ists.  Econom ics of E du-
cation R eview  7(2): 251±255.
Jordan, J. M ., M eador, M ., and Walters, S . J.  K . (1989). A cademic research productivity,
department size, and organization: F urther results.  Economics of Education R eview 
8(24): 345±352.
K yvik, S . (1995). A re big university departments better than sm all ones?  H igher Educa-
tion  30(3): 295±304.
Laband, D . N . (1985). A n evaluation of 50 ªrankedº economics departm ents Ðby quan-
tity and quality of faculty publication and graduate student productivity.  Southern E co-
nom ic Journal  52(1): 216 ±240.
Levin, S . G ., and S tephan, P. E. (1989). A ge and research productivity of academ ic
scientists.  Research in H igher Education  30(5): 531±549.

DETERM INANTS OF RESEARCH  PRODUCTIVITY IN  HIGHER EDUCATION631 
L ew is, D . R., and Becker, W. E. (eds.) (1979).  A cadem ic R ewards in Higher Education .
C am bridge, M A : Ballinger P ublishing.
L ew is, D . R., and D undar, H . (1995). Economies of scale and scope in Turkish univer-
sities.  E ducation E conom ics  3(2): 133±157.
L odahl, J. B ., and G ordon, G . (1972). T he structure of scientific fields and the function-
ing of university graduate departm ents.  A m erican Sociological R eview  37 (F ebruary):
57±72.
L ong, J.  S . (1978). P roductivity and academ ic positions in the scientific career.  A m erican
Sociological Review  43 (D ecem ber): 889±908.
M assey, W. F., and Wilger, A . K . (1995). Improving productivity: W hat faculty think
about it Ðand its effect on quality.  C hange  27(4): 1±20.
M eador, M ., Walters, S . J. K ., and Jordan, J.  M . (1992). A cadem ic research productivity. 
Econom ics of E ducation Review  11(2): 161±167.
O lson, J. E . (1994). Institutional and technical constraints on faculty gross productivity
in A merican doctoral universities.  Research in H igher E ducation  35(5): 549 ±567.
R oose, K. K ., and A nderson, C. J. (1970).  A  R ating of G raduate P rogram s . Washington,
D C : The A m erican C ouncil on H igher Education.
R ushton, J. P., and Meltzer, S . (1981). Research productivity, university revenue, and
scholarly im pact (citations) of 169 B ritish, C anadian, and U nited S tates universities. 
Scientometrics  3: 275±303.
S aunier, M. E . (1985). O bjective measures as  predictors of reputational ratings.  Research
in H igher Education  23(3): 227±244.
Tan, D . L . (1990). Is there a better w ay to m easure quality of program s? Lyon, F rance:
U npublished paper presented at the 1990 Annual C onference of the E uropean A ssocia-
tion for Institutional Research.
Tan, D . L . (1992). A  m ultivariate approach to the assessment of quality.  R esearch in
H igher Education  33(2): 205±226.
Tautkoushian, R . K., D undar, H ., and B ecker, W. E. (in press). The N ational Research
C ouncil graduate program  ratings: W hat are they m easuring?  R eview  of Higher Edu-
cation .
Tien, F. F., and B lackburn, R . T. (1996). F aculty rank system , research m otivation, and
faculty research productivity.  Journal of H igher E ducation  67(1): 2±22.
Webster, D . S . (1986).  Academ ic Q uality R ankings of Am erican C olleges and U niver-
sities . S pringfield, IL: Charles C T hom as.
Young, D . L ., B lackburn, R . T., and Conrad, C. F. (1987). R esearch note: D im ensions of
program  quality in regional universities.  A m erican E ducational R esearch Journal 
24(2): 319±323. 
Received A pril 10, 1997.